Before you evaluate what a new piece of software can do, find out how you would leave it. Three questions, answered before anyone signs up for a trial: can you export your data in a format another tool can open, such as a spreadsheet file; is the plan billed monthly, or only annually; and who, by name, would do the canceling? If any answer is “we don’t know,” that is the first thing to find out, not the last.
It sounds backwards, but it changes every decision after it. A tool you can leave in an afternoon deserves a quick, cheap trial. A tool that holds your customer records in a format nothing else reads, on an annual contract, deserves the full 90 days below before you commit to it.
Where most small businesses actually stand
Most small businesses are still at the start of this. The JPMorganChase Institute, analyzing small business bank accounts, found that about 17.7 percent of firms had adopted AI by the end of 2025, in line with the Census Bureau’s figure of 17.8 percent (JPMorganChase Institute). Among non-adopters in the survey data it cites, 33 percent worried about tool quality and 28 percent about legal or compliance issues (JPMorganChase Institute). The same research put the typical adopter’s spending at roughly $28 to $30 a month (JPMorganChase Institute).
Those figures are about AI tools, but the lesson applies to any new software. In our view the subscription is rarely the expensive part; the expensive part is the hours your team spends learning a tool and moving work into it, which is why the plan below spends more time after sign-up than before it. If you have not yet settled which barrier is holding you back, our breakdown of the AI adoption challenges small businesses still face gives each one a self-test.
Three gates, 90 days
The plan is three decision points. Each gate has pass conditions written down before you reach it, and a tool that fails a gate stops there. Writing the conditions first matters because by day 30, whoever set the tool up will want it to succeed.
Gate 1, day 0: is this worth a trial?
- One problem, in one sentence. “Customers wait two days for a quote” is a problem. “We need better software” is not. If you have more problems than time, deciding which one gets attention this quarter is the job of a one-day strategic planning retreat, not of a software demo.
- One number that shows the problem today. Hours a week spent retyping orders, missed calls a week, days to send an invoice. If you do not already know it, count it for a week before the trial starts.
- The exit answers. Export format, billing term, and who cancels.
- One daily user. The person who will actually use the tool every day, named, and asked before the trial whether they want it. If the owner is the only person who has seen the demo, the trial is testing the owner’s enthusiasm, not the tool.
Pass: all four are written down. For a solo owner the daily user is you, which makes the number more important, not less, because nobody else will notice if it does not move.
Gate 2, day 30: did it work on real work?
Run the trial on a small, reversible slice of real work, such as one job type, one location or one group of customers, where a mistake cannot reach a customer unchecked. Keep the old way running alongside it. If the vendor’s free trial is shorter than 30 days, pay for one month on monthly billing rather than rushing the test.
At day 30, answer three questions:
- Did the number from gate 1 move, and by how much?
- Does the daily user reach for the new tool without being reminded? If they still do the work the old way and then enter it twice, the tool has not been adopted, whatever its dashboard says.
- What broke? Every new tool breaks something: a report nobody can find, a step that now takes longer. List each one.
Pass: the number moved and the daily user wants to keep the tool. A tool that moved the number but that the team avoids goes back to gate 1 with a training question; it does not move forward. For setting up the side-by-side comparison, with your prediction written down before the test, see our process improvement playbook.
Gate 3, day 90: does it stay?
Days 31 to 90 are for the work that makes a tool permanent: moving the rest of the work across, switching off the old way, and connecting the new tool to the ones you already run, so that nobody types the same customer into two systems. At day 90, check four things:
- The old way is switched off, or there is a written reason it is not.
- The gate 1 number is still better than it was on day 0, not just better than on day 30.
- A second person can run the tool if the daily user is out for a week.
- You would buy it again today, at today’s price, knowing what you now know.
Pass all four and this is the moment to consider annual billing, not before. Fail, and cancel using the exit answers you wrote on day 0. Either way, write down what you expected on day 0 next to what happened; that record is what makes the next purchase better, and it is the same habit we argue for in scoring the last business case before writing the next one.
A worked example, hypothetical
What follows is an invented illustration, not a real business. A four-person heating and cooling company takes service calls by phone and tracks jobs on a whiteboard. The owner is looking at a field-service scheduling app. At gate 1, the problem sentence is “techs arrive without the job history.” The number is callbacks, meaning return visits to fix a first visit, counted over the week before the trial. The office manager, who will run the schedule every day, is named as the daily user and wants to try it. The app exports jobs to a spreadsheet file and bills monthly.
The trial covers maintenance visits only. At day 30, callbacks on those visits are down, the office manager has stopped updating the whiteboard for them, and the thing that broke is the weekly invoice run, which now needs information from two places. That becomes the gate 3 work: connect the app to the accounting software, move repair calls across, and retire the whiteboard. Nobody’s job disappears in this story. The office manager stops copying job notes by hand and spends that time on the phone with customers.
Connecting the new tool to the old ones
The step that most often stalls between day 30 and day 90 is the connection: getting a new order in one tool to create an invoice, a contact or a job in another. Many tools connect to each other directly, so check the vendor’s integrations page first. When they do not, a no-code automation tool fills the gap. Three real options, with prices read from each vendor’s pricing page today:
- Zapier has a free plan with 100 tasks a month and two-step workflows; its Professional plan starts at $19.99 a month billed annually, or $29.99 billed monthly, for 750 tasks (Zapier pricing).
- Make has a free plan with 1,000 credits a month; its Core plan is $12 a month for 10,000 credits, with a discount for annual billing (Make pricing).
- n8n offers a free self-hosted Community Edition for anyone comfortable running their own server; its hosted Starter plan is listed at €20 a month billed annually for 2,500 workflow executions (n8n pricing).
For a small business joining two or three tools, a free tier is often enough to prove the connection works during the trial. Remember that an automation tool is one more tool holding your data, so it goes through gate 1 as well.
Where an AI assistant helps
An assistant is useful for the writing that makes the gates stick: turning a vague complaint into a one-sentence problem, drafting the day-30 questions for the daily user, and listing every system a new tool will touch before the trial begins. The Technology Adoption Evaluator prompt at BusinessPrompter.com is built for that last job. It is listed under Innovation & Growth as a Pro prompt, with a single-prompt version and a five-step chain, and it describes its output as a technology integration plan with steps, timelines and resource allocation, written for operations managers. Use it to draft the gate 3 plan, then hold that plan to the gates above.
What an assistant cannot answer is the second question at gate 2. Whether the daily user actually wants the tool is something you learn by asking them, and by watching whether they still keep the old notebook.
Our view: annual billing is a day-90 decision
This section is opinion. Vendors commonly discount annual plans, and Zapier’s pricing shows the size of the gap: $19.99 a month billed annually against $29.99 billed monthly for the same entry tier (Zapier pricing). The saving is real, but you pay for it with the exit. Paying the monthly price for the first 90 days is, in our view, the cheapest insurance a small business can buy on new software, because it keeps gate 3 an honest decision instead of a sunk cost. Switch to annual once the tool has passed, and put the renewal date in the calendar the same day.
We would also drop any evaluation that ranks tools on a long feature checklist. Features you will not use in the first 90 days do not count. One problem, one number, one daily user: a tool that wins on those three and loses on the checklist is the better buy.
Frequently asked questions
What is a technology adoption plan?
A written plan for taking a new tool from trial to everyday use, with the decisions set in advance. For a small business it can be three gates: day 0 (is it worth a trial), day 30 (did it work on real work) and day 90 (does it stay), each with pass conditions written down before you reach it.
How should a small business evaluate new technology?
Start with how you would leave: data export, billing term and who cancels. Then write down one problem, one number that measures it and one daily user, and trial the tool for 30 days on a small, reversible slice of real work while the old way keeps running.
How much do small businesses spend on AI tools?
The JPMorganChase Institute found typical spending of roughly $28 to $30 a month per firm among small businesses using AI in 2025, and an adoption rate of about 17.7 percent by the end of that year (Source: JPMorganChase Institute).
One question outlasts the whole plan. On every renewal date, open the invoice and ask whether you would sign up again today, at that price, knowing what the tool actually does in your business. If the answer takes more than a few seconds, go back to gate 1.
